Practically speaking, this represents a level of work that is nearly equivalent to building the entire model from scratch. In a large model (with hundreds of rows, for example), this process could be very time-consuming and error-prone.
If such updates are required to be done regularly, it can therefore make sense to create a model that allows this to be done in the most effective way.
This can be done by creating a model which stores any historic data in a separate range and – for any particular period – builds the forecast calculations using:
- The forecasting assumptions, for periods when historical data is not present.
- Historical data, if it is present. That is, the forecast essential recalculates the historical data in this case, so that a single forecast line can represent both the historical values and the “genuine” forecast.